Direct EPR Detection of the Carbonate Radical Anion Produced from Peroxynitrite and Carbon Dioxide*

The biological effects of peroxynitrite have been recently considered to be largely dependent on its reaction with carbon dioxide, which is present in high concentrations in intra- and extracellular compartments. Peroxynitrite anion (ONOO−) reacts rapidly with carbon dioxide, forming an adduct, nitrosoperoxocarboxylate (ONOOCO2 −), whose decomposition has been proposed to produce reactive intermediates such as the carbonate radical (CO·̄3). Here, by the use of rapid mixing continuous flow electron paramagnetic resonance (EPR), we directly detected the carbonate radical in flow mixtures of peroxynitrite with bicarbonate-carbon dioxide over the pH range of 6–9. The radical was unambiguously identified by its EPR parameters (g = 2.0113; line width = 5.5 G) and by experiments with bicarbonate labeled with 13C. In this case, the singlet EPR signal obtained with 12C bicarbonate splits into the expected doublet because of 13C (a(13C)= 11.7 G). The singlet spectrum of the unlabeled radical was invariant between pH 6 and 9, confirming that in this pH range the detected radical is the carbonate radical anion (CO·̄3). Importantly, in addition to contributing to the understanding of nitrosoperoxocarboxylate decomposition pathways, this is the first report unambiguously demonstrating the formation of the carbonate radical anion at physiological pHs by direct EPR spectroscopy.

Peroxynitrite 1 is formed from the very fast reaction between nitric oxide and superoxide anion (k ϭ (6.7 Ϫ 19) ϫ 10 9 M Ϫ1 s Ϫ1 ) (see Reaction 1) (1,2). The compound is a potent oxidant that has been receiving increasing attention as a potential pathogenic mediator in human diseases and as a cellular toxin in host defense mechanisms against invading microorganisms (3)(4)(5)(6). At present, a significant part of the biological reactivity of peroxynitrite is ascribed to the adduct produced by its reaction with carbon dioxide (7)(8)(9)(10)(11)(12)(13). The peroxynitrite anion (ONOO Ϫ ), which is the predominant form at physiological pHs (pK a ϭ 6.8) (see reaction 2, Table II) (2,3), reacts fast with carbon dioxide (pH-independent k ϭ 5.8 ϫ 10 4 M Ϫ1 ⅐ s Ϫ1 at 37°C) (11), producing an adduct whose structure is proposed to be ([ONOOCO 2 ] Ϫ , nitrosoperoxocarboxylate) (see reaction 3, Table II) (7). Taking into account the concentrations of carbon dioxide in equilibrium with bicarbonate present in physiological fluids, model calculations have suggested that most of the peroxynitrite that might be formed in these fluids will produce the carbon dioxide adduct before reacting with other biological targets (5,13).
Carbon dioxide modulates the reactivity of peroxynitrite by altering reaction rates, product yields, and product distribution (7)(8)(9)(10)(11)(12)(13). In these reactions, formation of the adduct nitrosoperoxocarboxylate is rate-limiting, as first proposed by Lymar and Hurst (7). This suggestion was confirmed by other authors (8 -13), and the current proposal is that in the absence of substrates, the carbon dioxide adduct decomposes to nitrate and carbon dioxide, but in their presence, a fraction of the adduct (ϳ35%) can engage in one-electron oxidations (and aromatic nitrations), probably through homolysis to nitrogen dioxide and carbonate radical anion (Scheme 1A). The participation of the carbonate radical in the oxidation of substrates was inferred from the reactivity of peroxynitrite with inorganic ions in the presence of carbon dioxide (11,12).
The pathway shown in Scheme 1A is similar to the one initially proposed to explain the oxidations mediated by peroxynitrite in the absence of carbon dioxide (Scheme 1B). It has been proposed that peroxynitrous acid (ONOOH) promotes oneelectron oxidations following a rate-limiting unimolecular activation to a species whose chemical identity, an activated form of peroxynitrous acid (ONOOH*) or the hydroxyl radical, remained under debate for a long time (3, 14 -18). It was only recently that clear experimental evidence was obtained demonstrating that a significant portion of the oxidative activity of peroxynitrous acid is because of the hydroxyl radical (19,20). The controversy was difficult to resolve because spin trapping of the hydroxyl radical with DMPO, 2 the most unambiguous method for detecting the hydroxyl radical by EPR (21) 1 Unless otherwise specified, the terms peroxynitrite and bicarbonate are used to refer to the sum of peroxynitrite anion (ONOO Ϫ ) and peroxynitrous acid (ONOOH) and to the sum of CO 2 , H 2 CO 3 , HCO 3 Ϫ , and CO 3 2Ϫ , respectively. IUPAC-recommended names for peroxynitrite anion, peroxynitrous acid, nitrosoperoxocarboxylate (ONOOCO 2 Ϫ ), and nitric oxide are oxoperoxynitrate(-1), hydrogen oxoperoxynitrate, 1-carboxylato-2-nitrosodioxidane, and nitrogen monoxide, respectively. 2 The abbreviation used is: DMPO, 5,5-dimethyl-1-pyrroline-N-oxide. and spin adducts (15,19). A direct demonstration of the participation of the carbonate radical in oxidations mediated by peroxynitrite in the presence of carbon dioxide (Scheme 1A) could be anticipated to be similarly difficult.
There is only one report of spin trapping of the carbonate radical with nitromethane in irradiated aqueous solutions at pH 12 (22). At physiological pHs, the DMPO-carbon dioxide radical anion produced during irradiation of bicarbonate and DMPO solutions was incorrectly attributed to the carbonate radical adduct (23). We have tried to trap the carbonate radical in systems containing peroxynitrite and bicarbonate under different experimental conditions but did not succeed. Consequently, we considered it worth trying to detect the carbonate radical directly by continuous fast flow EPR of peroxynitrite and bicarbonate solutions. Our results provided unambiguous evidence for the formation of the carbonate radical in these mixtures.

EXPERIMENTAL PROCEDURES
Materials-All reagents were purchased from Merck, Aldrich, or Sigma and were analytical grade or better. Sodium bicarbonate labeled with 13 C (98%) was obtained from Isotec (Miamisburg, OH). Peroxynitrite was synthesized from sodium nitrite (0.6 M) and hydrogen peroxide (0.65 M) in a quenched-flow reactor (3,9); excess hydrogen peroxide was used to minimize nitrite contamination (24). To eliminate excess hydrogen peroxide, peroxynitrite was treated with manganese dioxide. Synthesized peroxynitrite contained low levels of contaminating hydrogen peroxide (Ͻ 1%) and nitrite (10 -30%) that were determined as described previously (2) by the titanyl method and by absorbance measurements at 354 nm (⑀ ϭ 24.6 M Ϫ1 cm Ϫ1 ), respectively. The concentration of peroxynitrite stock solutions was determined spectrophotometrically at 302 nm using an extinction coefficient of 1670 M Ϫ1 cm Ϫ1 (3,9). All solutions were prepared with distilled water treated in a Millipore Milli-Q system.
EPR Experiments-The EPR fast-flow spectra were recorded at room temperature (25 Ϯ 2°C) on a Bruker EMX spectrometer operating at 9.68 GHz and 100 KHz field modulation equipped with a Bruker ER 4117 D-MTV dielectric mixing resonator with a 9-mm distance between the mixing cell and the resonator center. Sodium bicarbonate solutions were prepared in 0.1-1 M phosphate buffer of various pHs and left undisturbed for 15 min to permit bicarbonate-carbon dioxide equilibrium (7); the pH of each solution was determined after bicarbonate addition and equilibration. Peroxynitrite solutions were prepared with water or 0.2 M dipotassium hydrogen phosphate to a final pH of 12. These solutions were transferred to 2 20-ml syringes mounted on a syringe infusion pump (Harvard apparatus pump 22). Most spectra were recorded 15 ms after mixing under a continuous flow of 7 ml/min. The dispensed mixtures were collected for pH measurements at the end of the experiments to detect changes caused by mixing with the alkaline stock solutions of peroxynitrite. In some experiments, stock solutions of bicarbonate and peroxynitrite were prepared in buffers bubbled with nitrogen for 20 min. The concentrations of the carbonate radical were estimated by double integration of their EPR spectra and comparison with a doubly integrated signal obtained under the same experimental conditions from 10 M 4-hydroxy-2,2,6,6-tetramethyl-1-piperidinyloxy radical (g ϭ 2.0056) (25), which was also used to calibrate the magnetic field.

RESULTS
Fast flow mixing of concentrated solutions of peroxynitrite (10 mM final concentration) and bicarbonate (50 mM final concentration) in 0.5 M phosphate buffer at pH 6.9 led to the detection of a one-line EPR spectrum (Fig. 1A) that was not detectable after stopping the flow (Fig. 1B). Also, no signal was detected in fast flow mixtures of the buffer with either bicarbonate (Fig. 1C) or peroxynitrite alone (Fig. 1D), indicating that the radical is produced from the reaction between peroxynitrite and carbon dioxide. Accordingly, the EPR signal varied significantly with the pH of the mixtures, and the intensities were maximum around pH 6.4 (Fig. 2). This value is similar to those obtained for maximum decomposition rates of peroxynitrite in the presence of carbon dioxide, which was in the pH range of 6.2-6.5 (7,9). These values are consistent with a rate-limiting reaction between the peroxynitrite anion (pK a ϭ SCHEME 1. FIG. 1. EPR continuous flow spectra of the carbonate radical produced from mixing peroxynitrite and bicarbonate solutions at room temperature. A, peroxynitrite (20 mM) and bicarbonate (100 mM) solutions in 1 M phosphate buffer, pH 6.9; B, same as A but with the spectrum scanned after stopping the flow; C, peroxynitrite (20 mM) mixed with 1 M phosphate buffer, pH 7.1; D, bicarbonate (100 mM) mixed with water, pH 12. The spectra were recorded 15 ms after mixing (flow of 7 ml/min) as described under "Experimental Procedures" except for the spectrum (B) that was scanned immediately after stopping the flow. The final concentrations of peroxynitrite, bicarbonate, and phosphate buffer were 10, 50, and 500 mM; the final pH was 6.9. Instrumental conditions: microwave power, 2 mW; modulation amplitude, 5G; time constant 0.16 s; scan rate 0.84 G/s. Each spectrum is the accumulation of two scans. 6.8) (3) and carbon dioxide (pK a apparent ϭ 6.4) (28) to produce the nitrosoperoxocarboxylate adduct whose homolysis would generate the carbonate radical (Scheme 1A).
To better characterize the radical detected during fast flow mixing of the peroxynitrite and bicarbonate solutions, we tried to increase the detectable radical concentrations under our instrumental conditions. As expected, the EPR signal shown in Fig. 1 increased with the flow in the 4 -8 ml/min range, but the available instrumentation only permitted flows of up to 8 ml/ min. The EPR signal did not change significantly when the buffer solutions were purged free of oxygen but grew when the concentrations of both peroxynitrite or bicarbonate were increased. Consequently, to obtain higher radical concentrations, we took advantage of the relatively slow carbon dioxide hydration-dehydration equilibrium as compared with peroxynitrite protonation and decay to nitrate (Scheme 1B) (7). For this purpose, solutions of 100 mM bicarbonate in 0.1 M phosphate buffer, pH 6.4, were fast mixed with alkaline solutions of 40 mM peroxynitrite to give solutions of pH 8.4 whose transient carbon dioxide concentrations (ϳ25 mM) were much higher than equilibrium values (ϳ1.23 mM). Under these conditions, the EPR signal obtained (Fig. 3A) was more intense than that obtained under conditions of equilibrium (Fig. 1A). Qualitatively, the signal was the same and showed EPR parameters, a line width of 5.5 G, and a g value of 2.0113, which were in good agreement with those previously reported for the carbonate radical in solid matrices if allowance is made for the very different media used, i.e. solutions versus solid matrix (29,30). Indeed, the matrix itself changes the EPR parameters of the carbonate radical, and some of the values reported in the literature are listed in Table I and compared with those determined here in aqueous solutions (Fig. 3A).
Unequivocal proof that the radical detected directly in fast flow mixtures of peroxynitrite with bicarbonate is the carbonate radical was obtained in experiments with bicarbonate labeled with 13 C (Fig. 3B). In this case, the singlet signal of the unlabeled carbonate radical (Fig. 3A) split into a half-height doublet signal (Fig. 3B) because of the contribution of the 13 C atom (I ϭ 1/2). The determined hyperfine splitting constant of 13 C is 11.7 G (Fig. 1B), a value in excellent agreement with those previously reported as experimental (Table I) (29,30) and calculated values (11 G) (31) for the labeled carbonate radical. The concentrations of the carbonate radical were estimated to be about 3 ϫ 10 Ϫ6 M under equilibrium conditions at pH 6.9 (Fig. 1A) and 5 ϫ 10 Ϫ6 M under nonequilibrium conditions at pH 8.4 (Fig. 3A). These values should be taken as rough estimates, because the experiments were run close to the detection limits of the instrument, and EPR itself presents inherent quantification problems (32). Still, the obtained values are of the same order of magnitude as those calculated by computer simulation considering reactions 2-10 (Table II) in which the decomposition of nitrosoperoxocarboxylate produces the carbonate radical anion with 35% yield (reaction 5a, Table II) and with k 4 assumed to be 1 ϫ 10 6 s Ϫ1 (7,17). In this scheme (Table  II), the carbonate radical anion was assumed to decay via three major pathways, including O . transfer to nitrogen dioxide (reaction 7), dismutation (reaction 8), and oxidation of residual

TABLE I EPR parameters of the carbonate radical in various media at 25°C
The parameters in aqueous solutions were determined from the spectra shown in Fig. 3, which were obtained from fast flow mixing of bicarbonate and peroxynitrite solutions as described in the legend to the figure. nitrite (reaction 9) (33)(34)(35). Under the same conditions as shown in Fig. 1A, the simulation showed that the transient concentration of the carbonate radical anion is 6 M at 15 ms (Fig. 4).
It is important to note that the one-line signal shown in Fig.  3A did not visibly split when the pH was changed over the 6 -9 range or by the use of modulation amplitudes as low as 1 G (data not shown). This indicates that either the proton hyperfine splitting constant of the carbonate radical is lower than 1 G or that the species is not protonated (see Reaction 11) over the studied pH range of 6 -9. Although the literature contains conflicting evidence regarding the protonated state of the carbonate radical over the 6 -12 pH range (36,37), recent studies by time resolved RAMAN spectroscopy have found no evidence for its protonation at pHs between 7.5 and 12.3 (38). These studies are in agreement with our EPR data indicating that the radical is not protonated at physiological pHs. CO 3 . ϩ H ϩ º HCO 3 .

DISCUSSION
A role for the carbonate radical in peroxynitrite-mediated processes was first suggested by Radi and co-workers (39), who demonstrated that bicarbonate greatly increased luminol chemiluminescence triggered by peroxynitrite. The direct reaction between peroxynitrite and carbon dioxide was then characterized by Lymar and Hurst (7). That this reaction should be a significant fraction of the biological reactivity of peroxynitrite and that it produces reactive oxidizing and nitrating species is currently supported by the work of many investigators (8 -13). The chemical nature of these reactive intermediates has also been proposed, but this is the first report presenting direct spectroscopic evidence for one of these intermediates, the carbonate radical anion (Figs 1 and 3, Table I). The proposed mechanisms of carbonate radical formation (reactions 3 and 5a) and consumption (reactions 7-9) ( Fig. 4; Table II) are consistent with the carbonate radical levels (ϳ3-5 ϫ 10 Ϫ6 M) quantified experimentally under equilibrium and nonequilibrium conditions.
Our results confirmed that the carbonate radical is not protonated at physiological pHs (Figs. 1-3) (38). As discussed by Bisby et al. (38), these results indicate that the carbonate radical anion is slightly more oxidizing in neutral solutions (E m,i ϳ 1.80 V) than previously assumed considering a pK a of 9.6 (E m,i ϳ 1.64 V) (38). Although local concentrations of carbon dioxide and biological targets will determine the decomposition pathway of peroxynitrite that might be formed in a particular physiological compartment, a significant fraction of peroxynitrite (35%) will produce reactive free radicals such as nitrogen dioxide and carbonate radical anion (Scheme 1A). Formation of hydroxyl radical is also possible (Scheme 1B) (19,20) but unlikely to be significant under physiological conditions, because the rate of peroxynitrite homolysis is too slow to efficiently compete with peroxynitrite attacking the many available biological targets (5,13). Both the hydroxyl radical and the carbonate radical anion can oxidize substrates by hydrogen atom abstraction, but in the case of the hydroxyl radical, the faster addition reactions would be likely to predominate in most conditions (40). This can explain why peroxynitrite-mediated nitration of aromatic amino acids is increased in the presence of carbon dioxide (5,13).
Direct EPR detection of the carbonate radical anion is important to unravel mechanistic details of oxidative damage inflicted not only by peroxynitrite but also by other oxidizing species such as hydrogen peroxide and the hydroxyl radical. On the one hand, normal plasma bicarbonate and carbon dioxide concentrations are 25 and 1.3 mM, but higher levels could be achieved during pathologic events such as respiratory distress syndrome or ischemia reperfusion (41). On the other hand, the carbonate radical has been proposed to be responsible for the effects of bicarbonate increasing photodamage to erythrocytes (42) and the bactericidal activity of the hydroxyl radical (5,23). The carbonate radical, however, has never been characterized in these systems. In a recent paper published by Hurst and co-workers (23), a role for the carbonate radical in the bactericidal effects of the hydroxyl radical is evidenced by the overall data. The claimed detection of the carbonate radical, however, was not substantiated by the presented spin trap results. The authors show that irradiated solutions of DMPO and bicarbonate presented a composite EPR spectrum whose components were the DMPO-hydroxyl radical adduct, the DMPO-hydrogen radical adduct, and a third adduct whose EPR parameters coincided with those of the DMPO-carbon dioxide radical adduct. This coincidence led the authors to conclude that they trapped the carbonate radical. However, in the carbon dioxide radical, most of the electron density is on the carbon atom,  (Table II) under the same conditions as those of the experiment in Fig. 1A. whereas in the carbonate radical, most of the electron density is on the oxygen atoms (30,38), and adducts of these radicals with DMPO are unlikely to have similar EPR parameters. In addition, to the best of our knowledge, nobody has been able to trap the carbonate radical with DMPO. There is a report of trapping of the carbonate radical with nitromethane by irradiation at alkaline pHs, and as expected, the nitromethanecarbonate radical adduct has EPR parameters similar to those of the hydroxyl radical (22), an oxygen-centered radical. The radical detected by Hurst and co-workers should, indeed, be the carbon dioxide radical anion produced from reduction of carbon dioxide by hydrogen atoms not completely quenched in their system; the presence of hydrogen atoms is proved by their detection of the DMPO-hydrogen radical adduct . In our experience, it is common to detect DMPO-carbon dioxide radical adduct from air-equilibrated buffer solutions in the presence of reducing agents such as hydrazine (43). CO  In summary, our results represent the first detection of the carbonate radical anion in aqueous solutions at physiological pHs. Detection and characterization of the radical as negatively charged at neutral pHs should contribute to the understanding of the roles of ubiquitous carbon dioxide in modulating the pathogenic mechanisms of peroxynitrite and other oxidizing intermediates.